Watch a repo get
explored in parallel.
A coordinator splits an unfamiliar codebase into modules. A subagent explores each one — in its own isolated context, at the same time — and their findings merge into one architecture map. Press run and watch it happen below.
Tested on a toy repo — and a real one
A tiny fixture is easy to get right. A widely-used open-source library the tool had no prior context on is the actual test.
Demo fixture
An intentionally small, self-contained codebase used to demo and test the pipeline end to end without needing a real project on hand. Its generated output is committed at examples/sample-output/ — readable with no setup.
Real-world test
A widely-used open-source library, chosen specifically because neither the tool nor its author had prior context on it. Output at examples/sample-output-click/ shows it adapting its own module-detection rule on the fly when click's source turned out to have no subdirectories — and flagging the deviation instead of forcing a bad fit.
Usage
No Claude Code install required just to read the committed example output.
# fastest way to see it work — just read the committed example cat examples/sample-output/ARCHITECTURE.md # run it yourself, against the included toy repo or any repo of your own bin/explain examples/toy-repo bin/explain /path/to/any/other/repo # or interactively, inside a Claude Code session /explain examples/toy-repo
Status
A weekend project built to learn Claude Code's subagent architecture hands-on — not a polished product.
pip install -r requirements.txt python scripts/validate_agents.py . # lint the skill/subagent config pytest tests/ -v # run the test suite # both run automatically in CI on every push